SpaceX and NVIDIA are preparing to move high-performance AI computing from terrestrial data centers into orbit. The companies plan to deploy a space-optimized version of NVIDIA’s Vera Rubin NVL72 system on SpaceX’s first-generation Starmind AI satellite, with the first launch targeted for the fourth quarter of 2027.
The Starmind program aims to place dedicated AI computing infrastructure in orbit. SpaceX says its AI satellites will capture solar energy and process workloads locally before sending results to Earth through high-bandwidth laser links connected to the Starlink network.
According to SpaceX, “Orbital computing may help solve some limitations of traditional data centers, such as power supply and cooling systems. The satellite will utilize the environment of space for its cooling and solar energy for power supply.”
SpaceX also plans to develop artificial intelligence satellites at the Gigasat Factory in Bastrop, Texas. SpaceX estimates that its Gigasat factory may produce thousands of AI satellites starting in 2027.
Space-based computing could allow satellites to process large volumes of sensor and imaging data before transmitting information to Earth. That could reduce downlink requirements and enable faster analysis for applications such as Earth observation, geospatial intelligence and autonomous space operations.
However, orbital AI remains technically demanding and capital-intensive. SpaceX and NVIDIA will need to demonstrate that the benefits of deploying advanced computing hardware in space outweigh the costs and complexity of launching, operating and upgrading the systems.
Also Read: Anthropic Reportedly Bans Salary, Equity Negotiations for New Hires